Using Forecasting Software to Drive Specific Business Improvements – Example One

Posted on June 26, 2013 by Gene Tanski

I am going to veer off from best practices for just a blog or two – we were recently going through a couple of project review sessions with clients and it struck me that there were some good examples of how to specifically apply improvements with forecasting software. So thought I would share them; and for a special treat, I happened to have been running through the trails by my home and for some reason the Beverly Hillbillies sprang into my mind so the opening paragraph – please read it to the opening stanza of the BH theme song:

Want to tell you a little story about some companies, looking to establish best practices in S&OP.

Then one day, as they were boarding on despair, they discovered the miracle of forecasting software.

Okay – that was lame but it made me giggle, and I have always enjoyed the incredible gift of cracking myself up.

The point here is that the specific improvement example is based on using the more accurate forecast from the forecasting software to drive performance improvements in both inventory and customer service.

Specifically, this example happened with two different companies in different types of business: one in paper and one in lawn care. So this type of work should have broad application.

Upon applying the best practice of measurement (please see earlier blog on S&OP software investment), these companies were able to identify opportunities with their top customers. One – they had indeed improved their forecasting in excess of 25%. In addition, they were able to show that the forecast supplied by their customers had error in excess of 50% (don’t think I need to bring up the dart board analogy here). Please remember that we are talking about forecast error measured at the execution level, which in one case is by SKU by customer and in the other is SKU by ship to location. Lastly, they were able to show what the implications were in terms of product inventory that was held throughout the value chain – specifically customer stores and distribution centers as well as our clients’ warehouses.

Incidentally, I should also point out that the companies were now running their computer forecasting based on the combined input of historical internal information (shipment, inventory, etc.) along with Point of Sale data from their customers (both sales at stores and inventory at stores and DCs).

Armed with this revealing information, these companies approached their customers. Based on the data and information that was provided (of course having gone through a verification process), the partners in the value chain were able to take specific action.

With regards to the paper company, they were able to work with their client to streamline the new product introduction process. Since they can sense the sales activity throughout the value chain, they can more specifically match the timing of introducing the new product upgrade to when the old product will sell out in the normal course of business. For the retail customer, they were able to cut out almost $2 million in working capital carrying unneeded inventory at their DCs and stores. Our client was able to save $1 million plus per quarter in promotion incentives used to clear out old product in anticipation of the new product intro.

Focusing on lawn care, our client’s retail partner was ordering far too much product in anticipation of the spring season, which took up space in DCs, required extra space in the stores, and far exceeded actual demand. While I do not have the exact figures, we do have verification that this unnecessarily consumed working capital and sub-optimized the performance of many of their stores as that retail space could be used for other products with equal demand. Our client in turn was able to avoid the inevitable return process and accompanying issue of product credits. In addition, they were able to utilize constrained capacity for a broader array of products, some with higher profit margin. They are estimating improved profitability of approximately 6% and counting. The counting part stems from the application of the new practices to their remaining base of retail partners.

The most accurate forecast possible derived from your forecasting software is worthless unless you apply it to specific components of your value chain and back it up through measurement-in other words, discover your “bubbling crude” and “Texas Tea”. Hopefully these two examples will provide the basis for some creative thinking within your value chain, and some measurable benefits to your bottom line.

Causal Factors in the Recent News – Using Them for Your Forecast?

Posted on January 31, 2013 by Gene Tanski

Great information in the news this week – I am sure everyone reads the news with a specific focus on how to improve forecasting and S&OP, right? Anyway – to the point.

Yields on 10 year treasuries rose 11 basis points. Claims for jobless benefits fell to a 5 year low. US home sales for the month of Dec. fell 7.3%, with a 1.3% increase in average sales price, but for year over year sales actually increased 8.8%, with a 14% rise in average sales price.

From an S&OP and demand planning/forecasting point of view, this information raises some interesting questions about causal impact and associated lead times.

What does all of this information imply for demand for your products 3 months from now, 6 months from now? For future pricing of your raw materials? Do the implications change per geography? Per season? Per product line or per customer class?

Does the fact that banks are accelerating repayment of bailout funds mean anything regarding interest rates? And do interest rates impact your procurement policies or demand for your products?

And what is the lead time associated with all of these potentially causal and leading indicators?

Most importantly, does your current process and technology platform allow you to study the impacts, understand the causality, plan for the impact with correct lead time? Are you having these conversations?

Our clients are talking about the implications of increased housing values on consumer spending and in what time frame. They are creating scenarios around different inflation trajectories, different employment rate scenarios and doing this within the context of different weather projections by geographic region. Are you equipped to have similar conversations on a weekly, monthly quarterly basis?

CPG and B2B manufacturing companies use leading indicators / causal factors like weather (temp. and precip. indexes), unemployment, birth rates, housing starts, and Nymex commodity spot prices to improve their forecasting and do so for cascading time frames. They are included in as additional inputs into forecasting process and used concurrently with the traditional historical inputs.

We recently had a good conversation with Amber Sally from Gartner. Among the topics discussed were her views on competitive differentiators in demand planning and forecasting – one was the ability to employ causal factors into your process.

Are you going to figure out how to corral the power of this competitive differentiator, or are you going to passively let your competition outflank you just as you are passively allowing external forces to impact your business without understanding the details of the how, why, where, and how much?

External Data – The Next Frontier

Posted on December 7, 2010 by Gene Tanski

Considering External Data in Demand Planning

When you make business decisions, you are almost always looking forward. You’re thinking about the future. This means you’re thinking about forecasting and planning. And what this really all means, is that you are thinking about demand.

When you make critical decisions around demand, you need information from disparate data sources about your market, your customers, the economy, the weather, your costs, your profit margins, what your competitors are doing, your ability to stimulate demand, your company’s capacity to meet the expected demand, among other things.

When thinking about all these factors, and when looking into the future to make these decisions and achieve the desired outcomes, it seems obvious that we have a problem. The data that has historically fed the forecast (past orders, past shipments, last month’s prices) is going to be less than adequate.

The fact of the matter is that the most important business decisions, and the fundamental practice of forecasting itself, are based primarily on backward-looking, after-the-fact, inward-looking data. This is why it’s no mystery that a majority of companies are dealing with 50% or greater error at the execution level of forecasting measured on an absolute basis. With this in mind, can you think of a better way to control business costs than to reduce forecasting error?

Besides improving the forecast platform, improved information would be the single most important tool for reducing forecasting error. If you could obtain more direct feedback from your customers (what they’re selling, what they have in inventory, how their latest promotion was performing) would that help your forecasting and planning? If your business is seasonal, would more accurate weather information help in your decision-making? If you supplied the residential construction market, would daily updates on housing starts have an impact? Does unemployment impact the consumption of high-end craft beers, and does this vary by region? Does overcapacity in the market among your competitors impact your profit margins? Do more than two or three external factors impact demand for your product at the same time?

Yup – we are going to have to talk more about this one. External data quality and timeliness, and then managing that information to optimize your decisions, will be the next frontier in business management. The good news is that Demand Foresight believes we have the platform, and there is now a great proliferation of more accurate information and competitive data markets available for more industries. Any input, experiences, examples, questions and critiques will be welcome.

What our forecasting and planning software can do.

Demand Planning blogosphere: Yokohama and BusinessWeek’s “Plan V”

Posted on July 17, 2010 by Gene Tanski

“Yokohama Tire Canada: Forecast Accuracy and the Cost of Being Right“

Offering some comment and reaction to supply chain, forecasting and demand planning blog posts that caught my eye in recent months. In April, Jonathon Karelse, marketing manager of Yokohama Tire, teased a then-upcoming presentation he was giving at an Institute of Business Forecasting event.

While I did not have the opportunity to see Jonathan’s presentation, one claim he makes in his preview post did catch my eye. I bring it up because I often find myself in some version of this discussion when I’m talking with companies that want to attack their supply chain management issues and become more profitable.

Consider that the result of demand planning is only profitable to the extent that it is actionable – that is, if links above or below in the participant’s supply chain are unable to respond to the data, it might be a purely academic exercise.

I contend that it is not academic at all — even if your supply chain can’t react — because the issue isn’t forecast, the issue is timing. If the forecast identifies a demand signal that a company can’t or won’t respond to, it’s still good to know in the first place. Secondly, perhaps next time the company can learn and adjust for the opportunity by adding extra capacity or additional storage in a more favorable location or what have you; again, tough to say without seeing the speech. Last, you can calculate the opportunity cost of not having acted.

Karelse correctly points out the importance of keeping cost and benefits in mind — no argument from me — but the business Intelligence that he mentions should be integrated into a company’s demand planning, S&OP and forecasting processes with the software being used — not as a parallel process. Upper management is absolutely correct to drive for utmost accuracy, because even if the company decides not to act on a forecast (e.g., someone is optimally supplied), at least they are able to measure the difference and understand the opportunity cost.

“Why Every Business Needs a ‘Plan V’”

A BusinessWeek guest blog from Harold L. Sirkin used the recent volcanic eruptions in Iceland as a jumping-off point for musings on modern-day continuity planningn for business in the global age:

We may or may not see more such disruptions; who knows? What we do know is that any disruption that does occur will have far more serious ripple effects than anything seen in the past. That’s what happens when companies from everywhere are competing for everything with companies from everyplace else.

Wouldn’t debate this any more than I would over apple pie being good; however, how do you move from the theoretical stance below to the potentially unsustainable practice of simultaneously being ready for everything from global warming to Godzilla? For manufacturers, he recommends strategic inventory reserves, redundant manufacturing locations and alternative distribution networks. I don’t know many mid-sized manufacturers who can afford such preparedness.

It is extremely hard to run a profitable business focused on the .1% scenario and still make enough money to be alive when the .1% scenario comes to pass. This isn’t to say that you shouldn’t do the planning. This is something that can be helped by a process (and supporting technology) that combines comprehensive short- and long-term (high-level and execution-level detail, as well) forecasting and planning scenarios, taking into account the day-to-day business and what is most profitable right now.

From this assumption, you can then allow management to consider some strategic options. For example, “It may cost us a bit more to have an extra DC in South Africa, but the sales forecast shows growth that would support that DC in about three years. Why don’t we accelerate that investment to provide more fleixibility now and provide competitive insurance in case the .1% scenario happens.” That may be a more practical and sustainable approach.

POS data and beyond: will your forecasting capacity be outstripped by the data gusher?

Posted on June 23, 2010 by Gene Tanski

I had the pleasure of speaking with supply chain VP Noha Tohamy of Gartner/AMR Research recently. We were discussing exotic new taxonomies of demand forecasting and demand planning — demand sensing in particular, and where it fits within our understanding of the S&OP process. We also talked about the value of a holistic approach to demand planning — something we discussed earlier on this blog. Noha agreed with the Foresight premise that a complete, holistic approach to demand planning and forecasting was best: the ability to look at short, medium and long time frames at multiple levels of detail (store/customer to entire networks and everything in between) all at the same time. However, she highlighted a couple of critical caveats. One was the assumption that a company has a platform to enable such a comprehensive approach. Enough said there.

The other caveat is data. In some instances, forecasting and analytical computing ability exceeds the timeliness and accuracy of the data available; in others, there is more data available than some technologies, processes and personnel are capable of handling. Often both these scenarios can be found at the same company. Which problem do you tackle first? Or can you take them both on at once?

Putting POS data in play: a field example

Let’s get a bit more detailed. In terms of importance for improving supply chain efficiency and building partnerships with your customers, POS (point of sale) data ranks high. But it also serves as a good example of the conundrum expressed above.

More advanced platforms and better collaboration have given supply chain technology vendors the ability to grab POS data that is made available either directly by the retail stores or through aggregators such as VIP or IRI. This data is often made available on a weekly basis and aggregated at a chain/region level rather than by store. Historically, the reasons for this have been data size, network connections etc., but these reasons are becoming less relevant as technology improves.

For POS data, even the weekly schedule is frequent enough to allow demand planning applications to grab it and leverage it for modeling, inventory management, trend reporting, and more. In fact, these same applications would allow for the same manipulation and value to be derived from more timely POS — say, on a daily basis and at a store level — which would allow suppliers to more actively respond to trends such as overselling in response to a promotion. Demand Foresight brings its clients this capability, as do others such as Terra Technologies.

Wal-Mart and Home Depot: stunning implications

However, now groups like Wal-Mart and Home Depot are advancing an even more immediate data model. They are actively prototyping a business model wherein they do not take ownership of the product at their stores until the transaction at the cash register — basically a back-to-back where the retailer collects cash from the customer, only having “owned” the product for a fraction of a second, then paying the supplier of the product 30, 45 or 60 days later.

There are several ramifications for this model, but let’s focus on the potentially stunning impact in data alone. How many of you (and your companies that supply retailers, or supply companies that supply retailers) have platformed themselves to take in huge amounts of POS data once a day? Four times a day? Hourly? Fractions of an hour? Given the model described above and the facts that:

• You have responsibility for the product until it is actually sold
• You are still operating under service level agreements that demand coverage levels for each square foot of shelf space

Are you prepared? Most are not and this is an example of data overtaking technical capabilities.

Broadening the data question

And remember, this is just one type of data. What about data associated with capacity and the ability to promise orders? If a customer calls and requests 10,000 pieces of product X, can customer service respond immediately with all the applicable availability, time and requirement answers? What if your products have a short shelf life? These are all critical to your forecast.

What about external data such as weather? Planalytics — a great company run by some outstanding ex-Air Force badasses — provides highly accurate and detailed weather info. Some are using the data to help with seasonality and impact on big sales days such as 4th of July. Now companies can profitably tackle formerly esoteric questions such as “How much more beer will we sell if the temp is 100 instead of 80?”

But even all this only scratches the surface of the minute and critical indicators available within Planalytics’ functionality. There’s still demographics, raw material pricing, market capacity…the list is endless, but also critical to answer as these indicators provide huge value for forecast accuracy in the short, medium and long term. Few applications are set up to fully take advantage of everything offered. This is the essence of the data conundrum.

So which side of the problem do you attack first to impact revenue in a positive way? Ideally, both. It will ever be our position that a holistic approach that marries detail with high-level strategy and the very short term with the long term is always going to provide the most value to an enterprise and its bottom line performance.

However, I think professionals are best served by being honest with what they really have to work with and then building from there — so if POS data is available on a weekly basis and it is not being used, then focus on taking advantage of the POS; learn the possibilities and build from there (invest in the correct applications). If however, you and POS are old friends, then tackle the other side, start to invest for the future by following tracks laid by companies like Best Buy and Sony, who are executing complete and real-time data sharing. If you platform yourself correctly, not only can you get ahead of the data curve, but you’ll be armed with a unique competitive weapon: the ability to approach your retail customers and say, “I know how you can optimize the dollars generated from each square foot of your stores and distribution centers.”

Hmmm – helping each partner of your value chain maximize their efficiency and profitability…Now there is something that can positively impact your EBITDA both for the long term and in a way that allows for competitive differentiation- and all of it tied to a focus on improving forecasting and demand planning.